Using Jev for chief.py task routing with a confidence threshold
Turns "which worker goes next" into a typed choice with a confidence score, routing to a human below 0.85, at about $0.42 per 10,000 decisions.
HOLY SH*T, YOUR CHIEF OF STAFF MAKES THIS EXACT DECISION 10,000 TIMES A DAY. "which worker goes next?" - asked hundreds of times, every single day. most teams still burn a frontier model on it. Jev answers it as a typed choice, with a confidence score attached: goal: compare 3 AI-agent tools choice: research confidence: 0.88 → above the 0.85 threshold cost: $0.000042 status: saved → queue/research drop below 0.85 and it doesn't guess - it goes to a human for review. no paragraph to parse, no coin flip. do the math: $0.042 per 1M input tokens, output free. 10,000 of these decisions cost $0.42. the LLM does the thinking. Jev decides where the work goes. code just executes. full setup + the http://chief.py router in the article ↓
This post quotes another post: x.com/0xkerazcity/status/2102042835352428590